Welcome to Stuff I Thought About Last Week, a personal collection of topics on tech, innovation, science, the digital economic transition, the finance industry, and whatever else made me think last week.

Click HERE to SIGN UP for SITALWeek’s Sunday Email.

In today’s post: Disney launches more adorable robots with potential commercial usage; new AI music tools are transformative to the art form; social media's shift away from news is leaving journalism in the lurch; copilots are the new UI and API layer for enterprise software; narrating your life; weather forecasting breakthrough; take a ride in Jeff Bezos' head; is consolidation the reason productivity is stalling in the US? And, much more below...

Stuff about Innovation and Technology
Emotive Bots
Disney’s adorable robots that I wrote about in #413 have already shown up in the “Star Wars: Galaxy Edge” theme park experience. Here is a video of the WALL-E-esque bots delighting guests. Disney used a combination of simulation, reinforcement learning, and 3D printing to rapidly produce bots that mimic personalities developed by character artists, going from concept to live, in-park deployment within a year. From the first US president audio-animatronic figures at the 1964 World’s Fair to these compelling bipedal character bots, Disney is likely one of the few top robotics and machine learning shops in the world, advantaged by their unique approach to robots. As one Disney Imagineer roboticist, Morgan Pope, put it: “We have a very different mission compared to conventional roboticists. We’re trying to use electromagnetism to create emotions.” Pope designs humanoid robots that can flip, jump, and swing around in stunt performances at the parks. Disney also has a new robot called Indestructible that rollerblades, somersaults, and does other stunts. I don’t know what Disney’s plans are beyond theme park attractions, but if you add an LLM to these bots, I think they could sell many millions to retail consumers. Disney’s robots also make me wonder if this more diminutively sized form factor could be of greater utility in commercial applications as well, rather than the full adult-sized bipedals most companies are working toward.

Music Is Dead. Long Live Music.
Google’s DeepMind introduced Lyria, a new set of tools to create music. Hum a few bars and you can transform a melody in your head into an instrumental or song. Another new feature called Dream Track allows users to describe a song they want in words and Lyria will generate it, including AI vocals from real singers like Charlie Puth or T-Pain. This music can then be set as the background for YouTube Shorts videos. While the fear of artists being completely replaced by generative AI is real, in the near term, I am looking forward to artists working with these tools to create new genres of music. The interesting business angle here is the musicians who are licensing their voices to the AI engines: is it a deal with the devil, or a leveraging of new technology? It’s pretty easy to envision how these advances will soon be applicable to video: describe a plot in detail, name your actors and director style, identify some music concepts, and you’ll have your own Hollywood studio. Here, longtime readers won't be surprised that I will yet again plug the 2014 movie The Congress for anyone who hasn’t seen it yet. Jeffrey Katzenberg says it will take 90% fewer people to make a big Hollywood animated movie thanks to AI. AI’s impact on the media industry is moving fast. Tom Hanks, commenting on future AI versions of himself, said: “some people are not going to dig it because it’s not really a real human being, and other people simply aren’t going to care because the story is OK. That’s not that different from any type of tool that [has] come into the moviemaking process since sound.”

No News Is Bad News
Recent shifts in user behavior and algorithm changes have caused traffic at news sources from social media to plummet by over 50% y/y and ~80% from 2020. The user behavior changes come from a shift toward social networking (i.e., TikTok) and video streaming that doesn’t emphasize linking to news/sources, while the algorithm change largely reflects Meta and X’s de-emphasis of news across their properties. It’s perhaps too early to declare the death of clickbait journalism, but it’s certainly taken some body blows and could be fatally wounded if algorithms and user behavior continue on their current trajectory. I believe there are other economic factors that will continue to shift users toward a smaller and smaller number of larger platforms, such as the steady increase in privacy changes on iOS and Android along with the pending loss of third-party cookies in Chrome. The logic is that larger platforms have enough first-party data to attract more advertisers at higher rates, while smaller ones will suffer a downward spiral in market share and ad dollars, rendering them unable to support their content. Media layoffs are up six-fold to ~20,000 people this year, according to Axios. Examples abound, such as Jezebel shutting down after sixteen years. 

Clickbait news was bad, but now people may increasingly get their news “reported” by influencers regurgitating distorted versions through their narrow (and often bonkers) lenses without linking to the original researched and edited news sources. Journalists don’t always get it right, but at least their accuracy rate is higher than that of some random, uncredentialed dude on TikTok. The loss in news traffic and monetization due to privacy changes will further cause journalism to focus on paying subscribers. For example, Insider is retreating back to just Business Insider, with co-founder Henry Blodget stepping down as CEO as the company pivots to focus “more emphasis on serving and attracting subscribers than generating traffic through social-media platforms.” The problem with news subscriptions is that it’s a rather niche business with limited revenues and reach, and, critically, it will exclude the broader group of folks used to having access to ad-supported news. This shift is likely to further polarize and fracture any sense of truth and common culture. I believe we need to reach a breaking point – where no one believes anything anymore – to reset culture, but it’s not clear how that endpoint will manifest, or whether we will reboot to something better or worse. In the meantime, expect an increasingly ignorant general population when it comes to current events.

Copilots Have Landed
GPT and LLM copilots built into operating systems are going to rapidly become the new user interface for consumers (see Discovery Engines), as well as enterprise software users. Imagine all-knowing AI that can talk to all of your enterprise apps, including WDAY, SAP, etc. Well, this completely new business UI is already here. Whereas apps have previously been able to speak to each other through APIs (as well as connective tissue linking programs and data created by large middleware layers like MuleSoft), going forward, custom GPT copilots will interact with workers and each other, creating a richer connection between apps and data. The apps themselves will increasingly retreat into the background as the UI changes from the app itself to a chatbot/copilot, which will shift proprietary value from software interfaces/functionality to exclusively data. Slack attempted such a transition by allowing queries of third-party apps from the console, but, with Microsoft’s Copilot Studio, I think we could see a new era of UI changes. While legacy cloud software will undoubtedly be around for a long time (just like legacy client/server software), its value is likely to be increasingly capped or transformed by AI. Further, once copilots/chatbots get the hang of things, humans will become marginalized in many of today’s popular business processes and workflows. If today’s large enterprise software companies cannot transition from user-based to value-based pricing and leverage proprietary pools of data, they face an uncertain future. 

SLMs
Microsoft has had success creating small language models. In contrast to LLMs, which are trained on a large set of real-world data and history, SLMs, like Microsoft’s Orca, are trained on tailored synthetic data. These lighter weight and lower power models could be deployed in a distributed manner, adding intelligence to a variety of things, like a butter-spreading bot. 

Querying Videos
Google’s Bard can now answer deeper questions about YouTube video content. For example: “if you’re looking for videos on how to make olive oil cake, you can now also ask how many eggs the recipe in the first video requires.” Training GPTs on videos will be key to giving them broader context and preparing them for real-world activities once they are embedded in robots. 

Synthetic Narration
A programmer used GPT-4V’s image processing capabilities and an AI voice simulator to create a live narration of screen grabs of his desktop as he worked, voiced by a synthetic David Attenborough. Add this to something like Meta’s Ray-Ban Smart Glasses and you could have your whole life narrated as you move through the world. Attenborough is none too excited about this development: "it is of the greatest concern to me that one day, and that day may now be very close, someone is going to use AI to deceive others into believing that I am saying things contrary to my beliefs or that misrepresent the wider concerns I have spent a lifetime trying to explain and promote." Luckily for Attenborough, in the not too distant future no one will believe anything is real.

WeatherAI
DeepMind beat the world’s leading weather forecasting model in 90% of cases using less energy than the supercomputers. Previously, I’ve joked that AI weather forecasting might use so much power that it would impact the climate itself, but this breakthrough proves otherwise. “We introduce GraphCast, a state-of-the-art AI model able to make medium-range weather forecasts with unprecedented accuracy. GraphCast predicts weather conditions up to 10 days in advance more accurately and much faster than the industry gold-standard weather simulation system – the High Resolution Forecast (HRES), produced by the European Centre for Medium-Range Weather Forecasts (ECMWF). GraphCast can also offer earlier warnings of extreme weather events. It can predict the tracks of cyclones with great accuracy further into the future, identifies atmospheric rivers associated with flood risk, and predicts the onset of extreme temperatures. This ability has the potential to save lives through greater preparedness.” Google is open sourcing the model.

Miscellaneous Stuff
Absurd Art
Artist Bobby Fingers makes a rowboat out of a giant Jeff Bezos head model. 

Boring Branding
Blockbuster drug names, like Zepbound, are weird in large part because they aren’t supposed to be interesting. 

Telescopic Timeline
The NYT has a great scroll-through of images from the JWST and a brief history of how our imaging of space-based objects has improved over the years.

Stuff About Demographics, the Economy, and Investing
M&A, PE Stifle Productivity and Innovation
Manufacturing productivity has been declining in the US for over a decade. Other Western countries, like Canada, the EU, and Britain, however, are all seeing rising productivity, albeit down materially from the earlier part of the 21st century, according to The Economist. After accounting for all sorts of potential explanations, The Economist mentions a few more speculative guesses. One that caught my attention was the lax antitrust M&A enforcement in the US. It sure seems like there is less competition in a lot of industries, and, in particular, PE-driven M&A has especially shifted corporate strategy to raising prices rather than growing profits through efficiency and innovation (see also The Circle of Debt). This seems to have created a complacency and lack of urgency in many industries to adopt or develop new technologies. That said, we appear to be on the cusp of a major productivity gain in manufacturing from AI-enabled robots, so this trend could reverse in spite of its mysterious sources.

✌️-Brad

Disclaimers:

The content of this newsletter is my personal opinion as of the date published and is subject to change without notice and may not reflect the opinion of NZS Capital, LLC.  This newsletter is an informal gathering of topics I’ve recently read and thought about. I will sometimes state things in the newsletter that contradict my own views in order to provoke debate. Often I try to make jokes, and they aren’t very funny – sorry. 

I may include links to third-party websites as a convenience, and the inclusion of such links does not imply any endorsement, approval, investigation, verification or monitoring by NZS Capital, LLC. If you choose to visit the linked sites, you do so at your own risk, and you will be subject to such sites' terms of use and privacy policies, over which NZS Capital, LLC has no control. In no event will NZS Capital, LLC be responsible for any information or content within the linked sites or your use of the linked sites.

Nothing in this newsletter should be construed as investment advice. The information contained herein is only as current as of the date indicated and may be superseded by subsequent market events or for other reasons. There is no guarantee that the information supplied is accurate, complete, or timely. Past performance is not a guarantee of future results. 

Investing involves risk, including the possible loss of principal and fluctuation of value. Nothing contained in this newsletter is an offer to sell or solicit any investment services or securities. Initial Public Offerings (IPOs) are highly speculative investments and may be subject to lower liquidity and greater volatility. Special risks associated with IPOs include limited operating history, unseasoned trading, high turnover and non-repeatable performance.